PhD Studentship: Marine Habitat Mapping Through Machine Learning

Project Title: Marine habitat mapping through machine learning

Marine habitats provide a wealth of ecosystem services to the marine and terrestrial environments and to people. At the same time, these habitats are under increasing pressure from human activities in the coastal zone and climate change impacts, leading to disturbance and degradation. Recognition of the ecosystem services that marine habitats provide has highlighted the need for conservation and restoration, however, such initiatives necessitate accurate mapping and monitoring of marine habitats at increasingly large spatial scales. 

Understanding what types of marine habitats exist in different locations and unravelling the underlying drivers of their distributions have great potential value for the overall protection and evaluation of ecosystem services. Habitat type drives biological community formation and ecosystem dynamics, with implications for biodiversity and carbon storage. Yet mapping marine habitats using traditional methods (e.g. in-situ observations by divers) is a costly and weather-dependent activity, and is further hindered by the large spatial scales involved. This highlights the need to develop new approaches and harness the increasing availability of sophisticated datasets and ability of machine learning techniques.  

This PhD project will utilise novel data for remote mapping of marine habitats in Southwest England, with opportunities to extend the approach that is developed more widely around the UK. Data sources to explore will include sonar, satellite/aerial imagery, acoustics, and others to be determined during the project. Sensor data will be combined with existing records of marine habitats to create training data for development of a machine learning tool to detect different habitat types. The project will first focus on the mapping of seagrass beds, which are known to provide invaluable ecosystem services but are undergoing worldwide loss and degradation. Once machine learning techniques and related workflows have been developed and tested in their ability to identify seagrass beds, the approach will be applied to other marine habitats (e.g. kelp forests, reef systems). In addition to habitat mapping, the project will investigate key factors that constrain the development of habitats in order to identify and prioritise areas for habitat rehabilitation.

The PhD student will work with an interdisciplinary research team involving geographers and data scientists from the University of Exeter and conservationists from lead partner, the Devon Wildlife Trust and the wider South West Wildlife Trust organisations. The student will join the CDT Environmental Intelligence at University of Exeter where they will receive training in relevant methods before conducting their research project. The project will make a critical contribution to the management of coastal systems by identifying areas of valuable marine habitats for protection and potential locations for restoration.

Advert information